Hi Humayun,
If I may intrude...
If the values you provided are in arcseconds then that would be VERY good.
But if these values are in pixels, then assuming 3.5"/pixel you are getting median FWHM of 6 arcseconds, Not as good as in the above scenario, but I suspect that when you are undersampling you will be naturally getting higher values of FWHM. Drizzle might help